Dermatology Cases with Medical Captions and Diagnostic Reasoning

7,041 dermatologic cases combine SkinCAP dermatologist-authored captions with SkinCoT clinician-certified diagnostic narratives. The dataset bridges the gap between visual observations and clinical reasoning. It was created by yuhos16 and last updated in October 2025.

Use Cases

• Train models to generate SkinCAP observation-first captions from dermatology images.
• Fine-tune language models to produce SkinCoT hierarchical diagnostic narratives based on image features.
• Develop multimodal models that jointly process images and text for diagnostic transparency.
• Benchmark model performance on the 7,041-case dataset for clinical reasoning tasks.

Strengths

• 7,041 total dermatologic cases provide a substantial foundation for training.
• Includes clinician-certified hierarchical Chain-of-Thought narratives for diagnostic reasoning.
• Captions are authored by dermatologists, ensuring clinical relevance.

Limitations

• Dataset size of 7,041 cases may be limited for training large-scale multimodal models.
• Geographic coverage is unknown, potentially limiting generalizability.
• Specific image modalities and resolutions are not detailed in the provided information.

Provenance

Source

huggingface dataset by yuhos16.

Collection Method

Combines dermatologist-authored captions (SkinCAP) with clinician-certified diagnostic narratives (SkinCoT).

Freshness

Last updated on 2025-10-27.